How to use lemma, POS and dependency of SpaCy in RASA NLU

[Swissistent (Patrick)]

Hello,

From SpaCy we would need the dependency and POS, and lemma in our extractor. Is there a also concept such as tokens or entities in rasa for dependencies, POS and lemma? What would be the best way to implement it? Either extending the spacy_tokenizer or creating another spacy_featurizer? Anyone who had that problems already?

Thanks, Patrick

[akelad (Akela Drissner)]

It’s probably best to create a new custom component, yes

[patrick (Patrick Nini)]

Thanks

[TatianaParshina]

You should create your custom tokenizer component based on tokenizer_spacy implementation.

I wrote post about it.